基于最优估计的数据融合理论  

The Theory of Data Fusion Based on State Optimal Estimation

作　　者：王炯琦[1] 周海银[1] 吴翊[1] 

机构地区：[1]国防科技大学理学院系统科学与数学系,湖南长沙410073

出　　处：《应用数学》2007年第2期392-399,共8页Mathematica Applicata

基　　金：国家自然科学基金资助项目(60272013);航天支撑技术基金项目(GFKD-HT-2006-02)

摘　　要：本文提出了一种最优加权的数据融合方法,分析了最优权值的分配原则;给出了多源信息统一的线性融合模型,使其表示不受数据类型和融合系统结构的限制,并指出在噪声协方差阵正定的前提下,线性最小方差估计融合和加权最小二乘估计融合是等价的;介绍了数据融合中的Bayes极大后验估计融合方法,给出了利用极大后验法进行传感器数据融合的一般表示公式;最后以两传感器数据融合为例,证明了利用Bayes极大后验估计进行两传感器数据融合所得到的融合状态的精度比相同条件下极大似然估计得到的精度要高,同时它们均优于任一单传感器局部估计精度.This paper advances a data fusion method with optimal weighted and analyzes the distribution principle for optimal weighted value ;presents an unified linear fusion model for multi-sensor information which can not be restricted by data types and fusion system structure ,moreover indicates that it is equivalent between linear minimum square estimation fusion and weighted least squares estimation fusion on premise of positive definite noise covariance matrix; introduces Bayes maximum verified estimation fusion method for data fusion and deduces general expression formula of sensor data fusion using maximum verified method; finally, takes data fusion for two sensors as an example to testify that the fusion state precision obtained by Bayes maximum verified estimation is much higher than that obtained by maximum likelihood estimation under the same condition; but both of them are much better than the local estimation precision for each single senor.

关 键 词：最优估计 数据融合 加权平均 BAYES估计 极大似然法 

分 类 号：TP212[自动化与计算机技术—检测技术与自动化装置]